Live MB OC Report :: ASUS Rampage Extreme

[JaVa_Az]

I also had random BSOD and a couple of RSOD using Vista 64. I did update all drivers for all hardware and it still did it. Moved ram inot other slot also the video card and it still would happen. I was so pissed so I tried to load Vista 32 and not a single BSOD. I have researching trying to firgure out why this is happened so untill then Ill run Vista 32.

I dont have the same ram but i do run some OCZ FLex XLC and i had kinda of the same problem in Windows XP64 bit. I could play games all night long run prime 95 for 3 plus hours, pass memtest. I personaly call that stable put a fork in it, It's done. But i would get freezes it was always at low power using the browser comming out of stand by. Sometimes i could go over 48 hours without it happening others it would happen 3 or 4 times in a couple hours. I tried just about every setting in the bios with no luck. I did some tests on the ROG speeding HDD so i wiped my drives for some write tests. When i reinstalled decided to go with xp sp3 32bit. Not one freeze since then. I am running the exact same bios settings as I had before with one change i turned off the marvell. But maybe it has something to do with 64bit drivers. I just did a cooling mod so i can get this q6600 to 4GHZ but i been kinda leary about changing any settings Since i seem to be 100% stable at the moment. I know a couple other people having trouble with OCZ i need to ask if they are running 64Bit i wonder if maybe that is the culprit

[zlojack]

I am running 64 bit and having some trouble. Still getting the hang of the board.

[RLM]

having just slapped an MCW30 directly on the X48 IHS

Just got my rampage extreme up and running. That's exactly what I've just done. That blocks been sat in a draw for about a year now. So the conclusion is though that the whole heatpipe assembly needs to go?

Got a lot of reading to do

RLM

[Zeus]

Yeah, the heatpipe looks nice but actually is a piece of junk cooling-wise.

Even with the washers in place and the swiftech block mounted onto the heatpipe assembly NB temps are still crap.

I will await Bitpower's new NB/SB block so i don't have to mess with small stick on heatsinks.

[RLM]

I will await Bitpower's new NB/SB block so i don't have to mess with small stick on heatsinks.

yeh, they look the business

RLM

[3oh6]

well, no comparison yet, just too damn busy. i did get the Noctua NC-U6 mounted and wouldn't you know it,
32M SPi at 600FSB tRD8 1.70vNB passed. here is a shot of the setup...



NB temp raised up to 49C during the run but the room is only 22C~23C so that definitely helps keep the temps
down. initial investigation appears to show the NC-U6, with active cooling, should have no problem with the
chipset directly mounted. for comparison, with the MCW30 mounted temps never go over 35C at those settings.

i am trying to find a couple MST-81s as i ran into a snag trying to run LN2 last night for the first time without
the stock HPA...the board is bowing like a smiley face in this setup and the board was acting like a crack head.
my plan is to run the MST-81s and get a local shop to make some aluminum back plates to mount on the back
side in hopes it helps keep the board flat. some stand offs on the socket mounting holes will also be fabbed i think.

[tekjunkie]

@EF, when you have your raid card on the board, does the message that says "Waiting for raid contoller firmware ro be ready" stay up during post for a while??. I just put mine up in the RE and the message during POST stays up for a while. This is with the Marvel Boot ROM disabled. Also, if anybody is interested, having the OCZ v2 on the Areca makes a world of difference, as opposed to having it on the ICH9R.

[RLM]

Memory and NB stability
-----------------------

To keep voltages at a min and just get the memory up to rated speed and timings, I worked up in 5mhz steps from 400-450fsb with 1:2, tuning up/down to min stable voltages at each stage. This took much too long but was still worth it for getting to know where I actually was. For that reason I've stuck with 0403 for now.

450fsb with 1:2, was unstable even at 1.45VNB and 1.95VDIMM with 100ps NB CLOCK SKEW delay (set manually to 100 at 410fsb).

Setting NB CLOCK SKEW to 200ps worked very well - at 450fsb, using AUTO / Normal / 100 / 300 gave me early freezes or errors in Prime or Intel (Linpack) and extra voltage did not help, but with 200, it behaves.

After 'finding' 200ps at 450fsb, I was able to lower VNB to 1.405V (min found) and VDIMM to 1.908v (min found) - both 0.05V lower (and stable) than where I was with 100ps (and unstable). 1.405VNB for PL7 @ 450/900/8-8-8-24 with 2x1GB is still a bit higher than I was hoping for though. DRAM SKEWS and NB GTL are currently on AUTO.

At these settings, ClockTwister causes instability if set to Strong or Stronger but at least the 200ps NB CLOCK SKEW allowed Moderate instead of Light.

Unfortunately, can't POST above 430fsb with PL6 and 1:2 (hangs on DET RAM) - tried VNB up to 1.48v + extra VDIMM but doesn't help. Will go back later and try sorting this but for now I'm going to push up from 450fsb keeping to 1:2 with PL7.

With PL7, Clocktwister on Moderate, 2N and slack 8-8-8-8-24, I'm losing memory performance all over the place: gives approx 10.5Gb/s Read with 50ns in Everest at this time.

Thanks to Tony at OCZ for providing some great info in his P5E3 Premium and X48 guides over there - I had a much better idea of when and why I might need to skew after finding those - I think a lot of the info is generally applicable to any X48 board and memory, not just the P5E3 Prem and OCZ.


CPU Voltage, temps and stability
-------------------------------

After the memory was stable at 900mhz (Orthos/IntelBurn(Linpack)/MemtestHCI/Memtestx86), I found the min CPUVCORE (LLC disabled) for each multi up to 9.5. 4275mhz (9.5) needs 1.33125V (set in BIOS) for 8 hour Prime. High for E8600, but my VID is 1.25 and this chip runs quite cool (Q820A606).

I set CPU clock skew to NORMAL at 400fsb but have not yet found benefit to either 100ps or 200ps delays for Vcore/stability.

Found no need (with dual-core) for extra CPUVTT or PLL at these levels - set manually to 1.10V and 1.510V (both min). Not expecting 450-500fsb to require extra CPUVTT or PLL either. GTLs all at AUTO.

CPU cooler is Noctua NH-U12P + 1 x NF-P12 fan. Hits 55c under Intel at 4275mhz/1.33125V, so it won't get that hot elsewhere. Have second fan for CPU cooler but not tried yet. Max temp shown in screen shot is Real Temp 2.70 which sets TJmax at 95c, which I think is correct (CoreTemp 99.3 auto sets TJmax at 100c but agrees with all temp changes)

I'm only using Intel for max CPU temps and quick full system tests (10xpass) - errors in here will always give me errors or worse in Prime, so it's quicker for rough stability. Surprisingly to me, condsidering the CPU temps generated by Intel, I found a 0.33125Vcore difference between Intel stable and Prime/Orthos stable with my board/CPU for 450x9.5 - Prime needs the extra Vcore or will restart/freeze/error on small FFTs. This Vcore difference showed the same in OCCT too (OCCT passes with 0.33125 less Vcore than Prime or Orthos). I'm not using LLC.

Nice post. Thanks for sharing.

Just done a bit of a search and I guess you are referring to this

http://www.ocztechnologyforum.com/fo...ad.php?t=41042

http://www.ocztechnologyforum.com/fo...ad.php?t=40746

RLM

[RLM]

i am trying to find a couple MST-81s as i ran into a snag trying to run LN2 last night for the first time without
the stock HPA...the board is bowing like a smiley face in this setup and the board was acting like a crack head.

interesting point. I've got mine mounted to the motherboard tray from a Lian Li. I cut a hole out the back for the backplate. That should do the trick shouldn't it?

RLM

[vega22]

I'm only using Intel for max CPU temps and quick full system tests (10xpass) - errors in here will always give me errors or worse in Prime, so it's quicker for rough stability. Surprisingly to me, condsidering the CPU temps generated by Intel, I found a 0.33125Vcore difference between Intel stable and Prime/Orthos stable with my board/CPU for 450x9.5 - Prime needs the extra Vcore or will restart/freeze/error on small FFTs. This Vcore difference showed the same in OCCT too (OCCT passes with 0.33125 less Vcore than Prime or Orthos). I'm not using LLC.

About the new Linpack in this board, i can say that is not enough to pass only 10xpass for to have a good idea about the stability (at least working with quads). My rig was fully stable @ 10xpass, even if I spent thousand 10xpass test. But it was unable to pass one only test @ 200xpass, i.e. It was drive me nut, and, moreover, the system became crazy in this type of test (100x+pass). So crazy that sometimes it was not enough to use the clear cmos, not even removing the battery, to restore a minimum of stability in the system (the Bios stability on this board is the worst I've seen in 15 years of experience, and it seems he does not know recover from serious failures without becoming corrupt).
Finally it seems that the value of VTT is one of the major keys to prevent such serious problems of stability. For 10xpass, i agree, for a little to moderate OC 1.1v is enough. But for more serious Linpack test i have to set the VTT @ 1.2v.

EDIT: all my test are in XP x64

[Johnny Bravo]

i am trying to find a couple MST-81s as i ran into a snag trying to run LN2 last night for the first time without
the stock HPA...the board is bowing like a smiley face in this setup and the board was acting like a crack head.
my plan is to run the MST-81s and get a local shop to make some aluminum back plates to mount on the back
side in hopes it helps keep the board flat. some stand offs on the socket mounting holes will also be fabbed i think.

Way ahead of you Jody - I'm using MST-81s, enzotech backplates and a few M3 screws and bolts to hold it all together - looks the business

[Zeus]

Way ahead of you Jody - I'm using MST-81s, enzotech backplates and a few M3 screws and bolts to hold it all together - looks the business

Since i'm planning to get those myself, i demand pics.

[runmc]

Way ahead of you Jody - I'm using MST-81s, enzotech backplates and a few M3 screws and bolts to hold it all together - looks the business

Yeah let's see what your cooling solution is Johnny. Pics please

My stock cooling is just not going to cut it either. I need to get it all replaced asap

[Johnny Bravo]

I'll get something sorted out ASAP

[rob2k]

Wow - new case and W/C parts have just arrived - spec in sig.

Time to pull my board apart. Could be some of these but hopefully more of these

[quattro_]

will the MST-81 fit both sides ? and will it work only using the push pin?

[cheapseats]

Might want to Try the newest real temp it is up to 2.79.4 Intel released the TjMax on 45nm Cpu`s and yours is a 100c TjMax

Thanks for the info Java - updated. I was 20 or so pages out of date on the RealTemp thread regarding Tjmax. And about 10 pages late in replying to your post - sorry

Nice post. Thanks for sharing.

Just done a bit of a search and I guess you are referring to this

http://www.ocztechnologyforum.com/fo...ad.php?t=41042

http://www.ocztechnologyforum.com/fo...ad.php?t=40746

RLM

Thanks RLM, those are the right links - forgot to include but I've added them now


Also found another error in my post:

I found a 0.33125Vcore difference between Intel stable and Prime/Orthos stable with my board/CPU for 450x9.5 - Prime needs the extra Vcore or will restart/freeze/error on small FFTs. This Vcore difference showed the same in OCCT too (OCCT passes with 0.33125 less Vcore than Prime or Orthos). I'm not using LLC.

the difference in 'stable' Vcore I found between Prime and Linpack is 0.03125V and not 0.33125V. I've updated that too

[Johnny Bravo]

As promised

[ToyTen]

nice

do you imroved results with this?and how far if possible ?

[runmc]

Yes that looks the business Johnny!! thanks

[eternal_fantasy]

@EF, when you have your raid card on the board, does the message that says "Waiting for raid contoller firmware ro be ready" stay up during post for a while??. I just put mine up in the RE and the message during POST stays up for a while. This is with the Marvel Boot ROM disabled. Also, if anybody is interested, having the OCZ v2 on the Areca makes a world of difference, as opposed to having it on the ICH9R.

Yeah that's the one drawback of using a RAID Controller card. Also having used the Adaptec card too, Areca takes the longest to initiate, around 20-30 secs.

The price you pay for performance...

[Zeus]

Those mosfetsinks look hawt! Will for sure go with that! Thanks Johnny.

[Pntgrd]

Those mosfetsinks look hawt! Will for sure go with that! Thanks Johnny.

Mine came in the mail today. They are actually heavier than I expected. I already have the little 4-pin ones and an Enzotech NB and SB cooler left over from a previous experiment. If I ever get around to actually trying out my RE I hope they all do the trick.

[3oh6]

As promised

excellent, thanks for the pics mate. now, do those backplates come with the MST-81s or are they something ordered separately...looks exactly like what I need. Just to find them in Canada now

interesting point. I've got mine mounted to the motherboard tray from a Lian Li. I cut a hole out the back for the backplate. That should do the trick shouldn't it?

RLM

you can see in some of my photos i use an MDF backboard with 1/2" thick insulation sheet in between the board and the back plate. this is what is causing the massive bowing i am seeing, i need some support for that area as there is no back plate. the mounting screws come right through the backboard then through the insulation sheet to the motherboard. i might experiment with a back plate in between the motherboard and the insulation to give myself the support it needs...we'll see. the MST-81s and backplates JB just showed off should do the trick as well because there wasn't this issue with the stock HPA. we'll see

[Zeus]

Don't think the backplates come together with the heatsinks but i guess the EK mosfet backplates will do just fine.